Read The Diligence Stack with Scottie
Keep The Diligence Stack. Lose the inbox pileup.
Scottie reads the newsletters you choose, keeps the useful differences, and attaches every original link. The Diligence Stack stays in your reading life without becoming another email to manage.
An independent guide. Scottie isn't affiliated with, endorsed by, or sponsored by The Diligence Stack. The publication owns its name, writing, and subscription terms.
From Scottie · July 31, 2026
See The Diligence Stack in a Scottie brief
Scottie read The Diligence Stack, SemiAnalysis, and Vik's Newsletter for a reader with the priorities shown below. Start with the rundown, open the full brief, or check every issue behind it.
3sources
6issues read
5stories included
3 of 3 The Diligence Stack issues included
What shaped this brief
Reader priorities
An infrastructure investor tracking hyperscaler capacity, AI storage, and the companies supplying data-center expansion.
These are illustrative priorities, not a customer’s data.
Sources in this brief
- The Diligence StackThe publication this guide is about
- SemiAnalysisAdds compute architecture, semiconductor economics, and detailed system costs.
- Vik's NewsletterAdds component-level analysis of power, memory, and optical interconnects.
Scottie
Scottie example brief
July 31, 2026 · Executive brief
01 / The rundown
- Utility power constraints are forcing hyperscalers toward leased shell landlords and onsite generation to overcome multi-year grid queues.
- Widespread trade labor shortages are driving data center builders toward modular designs, compressing construction schedules by over thirty percent.
- Compounding memory and packaging bottlenecks are lifting semiconductor average selling prices while spurring research into alternative memory architectures.
Read the complete brief 5 stories · 2 action items
02 / The briefing
01 / main
Power Bottlenecks Force Hyperscalers Into External Capacity Leases
The takeaway: Hyperscalers face growing backlog imbalances because grid studies take six to twelve months and substation work adds two to three years. To capture immediate demand, cloud giants increasingly rely on leased shells and external power developers.
Concrete details
- Backlog and RPO outpaced capex, pulling hyperscaler capacity coverage from 44% in 2024 down to 37% in 2026.
- Core Scientific billed 437 MW by mid-July against 1.1 GW leased, while Applied Digital delivered another 75 MW.
Why it matters for this reader: As an infrastructure investor tracking hyperscaler capacity, power bottlenecks highlight which physical landlords can secure terms without taking hardware risks.
Original sourcesThe Diligence Stack
02 / main
Onsite Generation Accelerates Data Center Power Without Grid Delays
The takeaway: Grid connection delays are making behind-the-meter generation a primary bridge for data center deployments. Shifting AI workloads toward inference lowers power spikes, making onsite power solutions far easier to deploy immediately.
Concrete details
- Bloom reported Q2 revenue of $1.065 billion, up 165.5% year-over-year, with GAAP gross margin reaching 33.4%.
- Interactive inference exhibits a maximum 9% power spike compared to 37.5% recorded during AI training workloads.
Why it matters for this reader: Evaluating behind-the-meter power options reveals which data-center expansion suppliers can help hyperscalers monetize capacity ahead of grid interconnections.
Original sourcesThe Diligence Stack
03 / main
Modular Construction Cuts Data Center Build Schedules By Months
The takeaway: Shortages of electricians and trade labor are pushing operators toward off-site prefabrication and modular data halls. Factory-built components lower capital expenditures per megawatt while cutting site delivery times by up to nine months.
Concrete details
- Modular construction compresses data center build windows by roughly 36% and reduces capex by 8% per megawatt.
- Electricians account for 30% to 40% of total construction hours, with severe trade labor shortages arriving by 2027.
Why it matters for this reader: Tracking modular construction suppliers helps you pinpoint which companies supplying data-center expansion can deliver capacity fastest amid severe trade labor shortages.
Original sourcesSemiAnalysis
04 / main
Expanding Hardware Bottlenecks Drive Higher Semiconductor Selling Prices
The takeaway: AI infrastructure forecasts underestimated how fast memory, packaging, and power needs would compound around accelerators. Semiconductor average selling prices are projected to nearly triple by 2028 as physical constraints dictate overall deployment speed.
Concrete details
- Current models project combined DRAM and NAND memory revenue to reach $800 billion to $850 billion by 2027.
- TSMC CoWoS capacity expectations increased from 30,000 monthly wafers to over 80,000 wafers per month for late 2025.
Why it matters for this reader: Monitoring supply chain bottlenecks beyond GPUs helps you assess capital spending efficiency across your hyperscaler capacity and hardware provider investments.
Original sourcesThe Diligence Stack
05 / main
High Bandwidth Memory Faces Severe Thermal And Wafer Efficiency Constraints
The takeaway: While HBM currently solves AI bandwidth demands, high wafer consumption and severe thermal issues are creating major manufacturing hurdles. Industry leaders expect new memory architectures to emerge as HBM packaging constraints increase costs.
Concrete details
- HBM die bit-density is one-third that of standard DRAM because through-silicon vias occupy physical chip space.
- HBM4 stacks sixteen dies for 64 GB capacity and 2 TB/s bandwidth, but requires three times more wafers.
Why it matters for this reader: Tracking AI storage and memory trade-offs helps you evaluate whether HBM wafer consumption will choke broader DRAM supply and impact server economics.
Original sourcesVik's Newsletter
Action items
- Analyze portfolio exposure to modular data center builders capable of easing trade labor shortages.
- Evaluate power equipment providers delivering behind-the-meter generation to hyperscalers facing grid delays.
See every issue behind this brief
Behind-the-Meter AI Buildout
- The Behind-the-Meter AI BuildoutThe Diligence Stack · Included
Hyperscaler Capacity Partner Hierarchy
- The Hyperscaler Capacity Partner HierarchyThe Diligence Stack · Included
Modular LEGO Datacenter Construction
- The Wild Wild West Of LEGO DatacentersSemiAnalysis · Included
Scaling Components Around AI Accelerators
- In 2023-2024 We Weren't Bullish EnoughThe Diligence Stack · Included
The Future of High Bandwidth Memory in AI
- HBM is the Ugly Baby: What Now?Vik's Newsletter · Included
AMD AI Software Progress
- Can AMD break the CUDA Moat? AMD Advancing AI 2026SemiAnalysis · Read, not included
About The Diligence Stack
Should you add The Diligence Stack to Scottie?
Follow AI infrastructure demand through capacity, storage, suppliers, and capital spending.[1][2][3]
Who it’s for
An infrastructure investor tracking hyperscaler capacity, AI storage, and the companies supplying data-center expansion.[1][2]
What you’ll find in it
The issues linked below include “The Agentic AI Storage Shock”, “The Hyperscaler Capacity Partner Hierarchy”, and “In 2023-2024 We Weren't Bullish Enough”. Open them to judge the publication in its own words.
- Byline
- By Creative Strategies[1]
How to read The Diligence Stack with Scottie
Add its public feed for publicly available issues. If you subscribe to paid issues, forward those emails to your private @scottie.fyi address. Scottie does not need access to your inbox.
Scottie read 3 The Diligence Stack items and included 3 for the brief. The sources covered different things, so Scottie kept their stories separate instead of forcing a connection.
Read three issues from The Diligence Stack
Read the publication in its own words. Scottie keeps these original links attached; it does not replace the writing.
- The Agentic AI Storage ShockPublic issue[2]
- The Hyperscaler Capacity Partner HierarchyPublic issue[3]
- In 2023-2024 We Weren't Bullish EnoughFri, 24 Jul 2026[4]